Hi !
On the analyzer i think i see the - 174dBm + 10log(RBW) + NF of S.A
for RBW=1KHz and NF of S.A = 30dB …Noice Input (analyzer with input at
50Ohm)=-114dBm
Is this right?
So today i did a simple experiment with an LNA . I saw that the output noice power was not input +GAIN + NF … i don’t know why this…
Check the attached please http://i1284.photobucket.com/albums/a57 … a75a76.png
Sorry, I am not following your descriptions or equations.
Can you expand to details please?
(FYI: spelling is Noise not noice)
Ηi waltr thanks for the comment about the noise!
So, the Displayed Average Noise level in dBm is: 10log(KT)+10log(RBW=IF filter of S.A)+ NF of S.A …
I think this is what i see in the display screen. Maye there is some also -( 2-3 dB) of noise that the Sample dETector adds…
Also i think the real display BW is 1.2*RBW ,because of gauss IF filter’s shape
So , the Signalout=Signalin (-90dBm)+Gain
Noiseout=Noisein+Gain+NFof LNA…
Ok, after some Google searches, reading and remembering my RF studies I think I am following your equations.
In: 10log(KT)+10log(RBW=IF filter of S.A)+ NF of S.A
The first term is “minimum equivalent input noise for the receiver” – 10log(k*T) = 10log( (1.38 x 10^-23 J/K * 290°K)*1000) = -174dBm/Hz
This assumes an ambient temperature of 290°K. If the temperature is different then the result is different.
the second term is ‘noise passed’ and is Bandwidth limited – 10log(BW) = 10log(1000Hz) = 30dB
Third is the NF of analyzer – NF = 30dB (this should be in the NA’s specs)
So, putting together the numbers: Noise level = -174dBm + 30dB + 30dB = -114dB. This looks very close to what you show in the first picture in your link, -112dBm is only 2dB off the calculation…
So now you add an Amplifier. So: -114dB + 3dB + 30dB = -81dB which is not what you see in the second picture so this equation does not work or some number are wrong.
What you need to do is calculate the Noise Figure for Cascaded Stages.
http://www.ece.ucsb.edu/yuegroup/Teachi … basics.pdf
http://www.dsplog.com/2012/04/21/noise- … ed-stages/
The Noise Factor of the SA = 1000, the LNA = 2
Gain of LNA = 30dB = 1000
Noise Factor = 2 + [(1000-1)/1000] = 2.999 (round to 3.0)
the Noise Figure is now 10log(3) = 4.77dB
What this means is that the SA has almost NO contribution to the overall noise figure and also shows the importance of a low NF in the first RF stage.
Now go back and combine these number into the first equation:
Noise level = -174dBm + 30 + 30 + 4.77 = -109dBm which is only a few dB different than your second picture.
Does this now make sense?
Not sure what you’re goal is, but a spectrum analyzer, with a low noise pre-amp, and narrow bandwidth (slow), won’t tell you much that is practical. For a real data radio, such as IEEE 802.15.4, there’s a 2MHz channel bandwidth. What you need to know is the power in that band (and a bit wider, as real radios do not have a channel-wide filter.
In these digital systems, the transmitter on to off ratio in time is huge, 1% or less is common. Depending on the use case.
So a spectrum analyzer needs to integrate the heck of of such channel bandwidth to even detect these fleeting signals.
Better yet, a real time spectrum analyzer ($75K) can run firmware that detects/demodulates a specific waveform, such as the DSSS with Offset QPSK that 802.15.4 uses.
Or the 20MHz channel of 802.11b/g and slow forms of 11n’s adaptive modulation.
The real benefit is using a real radio to measure the coherent noise power bandwidth. Non-coherent “noise” affects the SINR for the receiver. In spread spectrum, one gets the benefit of post-detection correlation gain. Also, the positive and negative if multipath is a big consideration. It’s quite statistic for a variety of channel conditions - such as non-line-of-sight in urban, low antenna heights, vs other conditions. See the IEEE 802.16 studies on the channel models.
So the point of all of this is that the Gaussian or thermal noise floor isn’t meaningful in modern spread spectrum and OFDM systems for data.
waltr:
Ok, after some Google searches, reading and remembering my RF studies I think I am following your equations.In: 10log(KT)+10log(RBW=IF filter of S.A)+ NF of S.A
The first term is “minimum equivalent input noise for the receiver” – 10log(k*T) = 10log( (1.38 x 10^-23 J/K * 290°K)*1000) = -174dBm/Hz
This assumes an ambient temperature of 290°K. If the temperature is different then the result is different.
the second term is ‘noise passed’ and is Bandwidth limited – 10log(BW) = 10log(1000Hz) = 30dB
Third is the NF of analyzer – NF = 30dB (this should be in the NA’s specs)
So, putting together the numbers: Noise level = -174dBm + 30dB + 30dB = -114dB. This looks very close to what you show in the first picture in your link, -112dBm is only 2dB off the calculation…
So now you add an Amplifier. So: -114dB + 3dB + 30dB = -81dB which is not what you see in the second picture so this equation does not work or some number are wrong.
What you need to do is calculate the Noise Figure for Cascaded Stages.
http://www.ece.ucsb.edu/yuegroup/Teachi … basics.pdf
http://www.dsplog.com/2012/04/21/noise- … ed-stages/
The Noise Factor of the SA = 1000, the LNA = 2
Gain of LNA = 30dB = 1000
Noise Factor = 2 + [(1000-1)/1000] = 2.999 (round to 3.0)
the Noise Figure is now 10log(3) = 4.77dB
What this means is that the SA has almost NO contribution to the overall noise figure and also shows the importance of a low NF in the first RF stage.
Now go back and combine these number into the first equation:
Noise level = -174dBm + 30 + 30 + 4.77 = -109dBm which is only a few dB different than your second picture.
Does this now make sense?
Yes yes many thanks…wow! It is cascade of course :shock: 
About the TotaL Noice Figure it is ok 4.77dB. So, Pout=KTBWGtotalNFtotal, so in terms of dB=ktbw+Gtotal+NFtotal
KTBW=-174dBm+30 (from RBW)
NFtotal is ok 4.77dB…
Gtotal=Glna+Gs.a right?
i miss the Gain of S.A at the equation i think :geek:
stevech:
Not sure what you’re goal is, but a spectrum analyzer, with a low noise pre-amp, and narrow bandwidth (slow), won’t tell you much that is practical. For a real data radio, such as IEEE 802.15.4, there’s a 2MHz channel bandwidth. What you need to know is the power in that band (and a bit wider, as real radios do not have a channel-wide filter.In these digital systems, the transmitter on to off ratio in time is huge, 1% or less is common. Depending on the use case.
So a spectrum analyzer needs to integrate the heck of of such channel bandwidth to even detect these fleeting signals.
Better yet, a real time spectrum analyzer ($75K) can run firmware that detects/demodulates a specific waveform, such as the DSSS with Offset QPSK that 802.15.4 uses.
Or the 20MHz channel of 802.11b/g and slow forms of 11n’s adaptive modulation.
The real benefit is using a real radio to measure the coherent noise power bandwidth. Non-coherent “noise” affects the SINR for the receiver. In spread spectrum, one gets the benefit of post-detection correlation gain. Also, the positive and negative if multipath is a big consideration. It’s quite statistic for a variety of channel conditions - such as non-line-of-sight in urban, low antenna heights, vs other conditions. See the IEEE 802.16 studies on the channel models.
So the point of all of this is that the Gaussian or thermal noise floor isn’t meaningful in modern spread spectrum and OFDM systems for data.
Hi. I just did a test with S.A to confirm if the displayed noise power come from S.A and LNA or from LNA alone.
one question. I have a transmitter sending 16QAM at 2465MHz BW=20KHz. The receiver some meters near take these data. So…
** to measure the real noise power input, i must put the S.A RBW equal to BW of my signal right? To detect how much noise is in the region of my signal?
So i will have: -174dBm+10logRBW=20KHz +NFs.a.
The real noise power input is the above without the NFs.a .i am right?
Sorry, I don’t know how to measure an LNA’s noise figure with just a spectrum analyzer. Involves coax lengths, etc. Once, I did measure a WiFi bi-directional amplifier’s LNA benefit to an 801.11 non-OFDM signal. The affordable LNA within the BDA was a negative benefit. These WiFi BDAs try to use an LNA to avoid a seriously unbalanced link (TX range far greater than RX range).
OFDM mode is a totally different issue than single carrier (among the 802.11 standards). It’s due to the high peak/average power ratio with OFDM - and that requires highly linear amplifiers, esp. the transmitting amplifier. Typical low cost amp has to back-off 6dB when going into OFDM modes. (seldom admitted to in WiFi marketing).
The LNA adds non-coherent noise - lots of it. It’s important in the receiver’s bandwidth, say 20MHz. So you could assess a fraction of that bandwidth in order to cope with the spectrum analyzer’s high noise floor. But I think special equipment is needed to truly get the noise figure. What really matters is the BER and FER for DSSS and OFDM modes at different SINRs, for a real system with an without the LNA. The LNA needs to be right at the antenna, not on the far end of coax.
In cellular base stations, and military and other special systems, there are some 1dB or so NF LNAs that do have benefit, but they use supercooled components are are very expensive.
stevech:
In cellular base stations, and military and other special systems, there are some 1dB or so NF LNAs that do have benefit, but they use supercooled components are are very expensive.
Room temperature, sub 1 dB NF LNA’s are now common and a complete amp can be built for under $100.
Take a look at http://www.hep.ucl.ac.uk/lc/T474/Docume … figure.pdf
It is quite possible to use an LNA to increase the dynamic range of the SA for noise measurements (or to “detect” signals near the noise).
